Розпізнавання зображень з попередньою нормалізацією геометричних перетворень на базі аналізу характерних точок

Abstract

The given work is devoted to solution of image recognition problem. Special attention is paid to the methods based on the analysis of characteristic points, which are determined by calculating descriptors. In the work, characteristic points are proposed to be used for the normalization of geometric transformations, which distinguish an input image from a template. To research this approach, the series of experiments was conducted: a matrix of geometric transformation, based on the analysis of SURF descriptors, was found and normalization was made. The research, carried out in this paper, has shown the feasibility of using characteristic points for the normalization of geometric transformations.